perspectives While the combined power of molecular and genetic analysis has led to major progress in our understanding of organisms from bacteria to Drosophila melanogaster, this combination of techniques has been very diffi- cult to apply to rnanm~s. A major part of the difficulty is the large size of the man~:~alian genome and the abundance of repetitive sequences. While in Drosophila both genetic and cytogenetic (hybridization to polytene chromosomes) analy- sis can resolve distances compa- rable to the length of DNA reco- vered in single X or cosmid clones, these techniques in

TIG --July 1986

Jumping libraries and linking libraries: the next generation of molecular tools in mammalian

genetics Annemarie Poustka and Hans Lehrach

New types of clone libraries (chromosome jumping libraries and IM~i.tJg !~rfaK~s) are bang developed to cope ~th the large physical distances separating markers and genes in mammalian genomes. Some a@ects of the technology are in the early stages of develOmmt, and a number of Moblems remain to be overcome. Necerthdess, we have successfully constructed a number of different jumln'ng and linking libraK~s from the human and mouse &enomes, and toe are beginning to use these long-range cloning and

nuCCn'ng techniques for many applications in mammalian gemelics.

mammals will only resolve or provide markers at distances of millions of hasepairs, at least 20 times the capacity of the larges," capacity cloning vectors (Fig. 1).

To close this gap between the resolution of mam- malian genetics or cT:ogenetics and the range easily covered by mole~ar cloning and DNA analysis, a number of laboratories have been working on the development of techniques able to cope with the dis- tances between genes in mmnmalian genomes ~. The approaches considered fall into two basic categories: serial techniques (chromosome walking and chromo- some jumping), and approaches allowing the parallel analysis of chromosome sub-regions, entire chromo- somes or genomes (linking clone analysis and establish- ment of ordered cosmid libraries).

Chromosome jumping At present the capacity of the available cloning vectors

limits the distances that can be analysed easily by serial

Gene complexes I ~ I

In situ:hvbridization

__genetics

Mammalian genome

Genetic mapping ]

l ! IGenos i

1 [_i Jumping :]

Molecular cloning [

i PFG gel blots 1 [ S°uthern I

blots

i i 10 4 10 ~

I

0.1

' I I i I

106 107 108 10 9 bp

| I I I

1 10 100 1000 cM Fig. 1, A comparison of the molecular (~) and genetic (cM) distances that can be covered in mammalian genomes by molecular and genetic approaches.

174

(walking) techniques. Chromosome walking, that is the repeated use of fragments from the end of a clone to identify adjacent overlapping clones, only proceeds in steps of less (and often much less~ ~a~i the average length of an insert. Although possib'~ities for the cloning of larger DNA fragments in other vector systems have been discussed, essentially all cloning of long stretches of DNA is carried out either in cosmid or ~. replacement vectors, limiting the insert sizes to less than 50 or less than 24 kb, respectively. If cosmid clones are used, a hypothetical chromosome walk of a million base pairs (I cM in the human genome) will therefore require, on a purely arithmetic basis, at least 20 and probably more than 40 steps, making this approach essentially unusable for these or larger distances. In addition, there are reasons for assuming that some mammalinn sequences might not be clonable in E. cob ~, thus terminating any chromosome walks across such a region.

Chromosome jumping, in contrast, aims to clone only the ends of large DNA fragments, while the rest of the DNA is deleted by a series of biochemical manipulations carried out before the cloning step. Although the length of the jumping clone insert, the sum of the two end fragments, is still limited by the capacity of the cloning vector, the distance between the two fragments can be hundreds of kilobases. In this approach only the end sequences of the original fragment, which constitute the final jumping clone, will have to be clonable or non- repetitive. This approach should therefore be less sensitive to sequences which are either unclonable or too repetitive to be crossed by chromosome walking.

The principle of this type of approach, which has been under development both in our laboratory and the laboratories of Collins and Weissman, is described in Fig. 2 3'4. Long DNA, prepared by lysis of cells imbedded in blocks of low melting-point agarose °, is digested very lightly within the blocks with, for example, BamHl. Partial digestion products are then separated by pulsed field gradient gel electrophoresis s and DNA is isolated from slices containing specific size cuts ranging up to hundreds of kilohases.

In order to permit the subsequent selective deletion of the middle of these large DNA fragments without concomitant random reassortment of end fragments, the DNAs are then circularized at very low DNA concentra- tion, usually in the presence of a plasmid sequence carrying a suppressor gene or a suppressor gene

© ~s, E~,~ s d ~ ~b~e~ S.V.. ~ t e ~ 0~- ~ m O . ~

TIG ~-- July 1986

High molecular weight DNA

digestion Size separation

Cyclization in presence of

"" marker plasmid

RI Bam

t Recut with _second t enzyme

t Ligate into t X vector

R R B B I ,.,~.- * ~ o .

F~. 2. Scl~wgc drawing of the conslnggon of co~lemmtao jumping libra6es: le~, DNA cleaved parlially with BamHl and fe.clmve.d with EmRI; 68M, DNA cleaved partially with EcoRl and m-cleaved with BamHl.

fragment 7'e. In a fraction of the molecules the suppres- sor gene will be included during circMarization and can be used later as a 'tag' to isolate or recognize the fragment contmning the junction between the two ends.

The large circular molecules are then cleaved into small fragments by complete digestion with a second enzyme (e.g. EcoRI) which does not cleave within the chosen suppressor plasmid. The resulting fragments are ligated into a ~. vector c a n 3 ~ amber mutations. Clones of the ligation product of the two ends of the large DNA fragments, identifiable by the suppressor-containing sequence, are selected by plating on bacterial hosts not can'ying a suppressor gene.

Clones identified by hybridization in a h'orary of

pe++spectives jumping fragments should therefore be derived from both ends of the original large DNA fragment after deletion of all sequences between the outermost sites for the enzyme used for the second cleavage.

A number of factors have to be taken into account in o p ' ~ the conditions for each step of the procedure. In particular, conditions favouring intramo~ecnbr riga- tion over intermolecular ligat~,~m have to be chosen for the circularization step. The ratio between circle forma- tion and intermoleculm" ligation events is governed.by two parameters: the effective loc~ concentration O) of one end of a molecule experienced by the other end of the same molecule, and the concentration (0 of the ends of all other long DNA molecules, determined by their molar concentration. The parameter j can be derived from the Jacobson-Stockmayer equationg:

~ . ~ x !0 4 j ' - kb3/2 M

as the equivalent concentration of linear

63.4 Y= ~m- pg mr-'

where kb is the length of the molecule in kilohase pair~.,. The fraction of intramolecular ligation events is t'Jen given by the ratio j/(i+j). These predictions have ~ e n verified experimentally in a model system 3. With increasing dilution, the ratio will asymptotically approach one, while the fraction of intermolecular events will approach but never reach zero. Therefore circu- larization has to be carded out at low DNA concentration. So that the sequence used as a tag can also be incorporated with a reasonable probabiliW during the circulat~tion step, the molar concentration of the tag has to be chosen to be roughly equivalent to j.

Since ligation in dilute solution will reduce, but not eliminate intermolecular lil~tion, prospective jumping clones v ~ usually have to be checked, for example by hybridization to appropriate panels of somatic cell hybrids or to pulsed field gradient gels, to ensure that both ends of the jumping clone come from the same large DNA fragment. The fraction of wrong clones can be considerably higher than expected statistically if, especially in attempts to jump very long distances, a significant fraction of the large DNA is broken and therefore unable to circu~rize, w~e still contffouting to the background of intermolecubr iigation events.

As illustrated in Fig. 3, chromosome jumping as d e ~ here is inherently directional. This is because

Fig. 3. Directional jumping R B R using two comptementmy

~ s ~ in Fig. 2. B,B~mffi!

/; Genomic DNA . B R RB B R F'-I ~ ~._ m

L i b ~

BFIB 175

perspectives High molecular weight DNA

JUMPING

• N o t l complete digestion

m J

C LINKING

Sau3A P2artial digestio~

Okb size cut

Dilute and cyclize t in presence i

of plasmid marker

©- ©-©

Recut BamHI ~ Recut Notl

© ;;- Q

t Ligate into X vector

B B N N

Fig. 4. Schematic drawing of the construction of r~r¢ cutter jamping libraries (lefO and complementary linkin# libraK, s

an EcoRI-BamHl fragment used as a probe will only identify clones derived from partial BamHI fragments extending in the direction of the EcoRl site. In a complementary library, constructed from ~gments generated by partial digestion withEcoRl ann re-cleaved with BamHl, only clones extending from the EcoRI site in the direction of the BamHl site should be identified. A clone identified in the complementary library by the end of the previous jump will therefore inherently extend in the same direction. Alternating between complement- ary libraries should give clones extending further and further in one or the other direction.

As in chromosome walking experiments, libraries of more than three genome equivalents (e.g. -,- 3 million clones for a BamHl partial library) have to be used to give a greater than 95% probability of finding a clone for the next step m. In practice, even larger libraries will be preferable, since a significant fraction of the identified end points will contain repetitive sequences. Clones with highly repetitive end points could be detected by hybridization with radioactively labelled total DNA ix, and thus at least these clones could be excluded from further analysis.

Rm-e cutter jumping libraries With the aim of reducing the otherwise very great

complexity of the libraries to be constructed and screened, our laboratory has especially concentrated on the use of enzymes, such as Notl, that cut rarely in

176

7"IC, - - July 1986

mammalian DNA (Fig. 4). While a library constructed from DNA fragments generated by partial digestion with an enzyme such as BamHl will typically have to contain millions of dones to be representative (requiring the ligation and packaging of the equivalent of hundreds of millions of clones) we expect that a library of a few (perhaps 3-10) thousand clones will contain most (don- able) Notl jumping clones (assuming that Notl fragments are on average a million bases long). In addition, the use of r~re cutting sites as start points and/or end points of chromosome jumps allows the distance covered by the jumping clone to be estimated by hybridizing the end ~jgments to blots of pulsed field gradient gels. How- ever, such an estimate has to be checked, since a jumping clone could end in a restriction site that might be cut only very infrequently in a digest of the whole genome (for example, bemuse of methylation of the DNA); if this were the case then a hybridizing band larger than the true jump size would be observed. One way of checking the estimate would be to analyse double-digest hybridizations.

Rare cutter libraries will be a very powerful tool for many applications, as long as the appropriate end fragments can be obtained. One difficulty is that prob- lems can be expected in circularizing the DNA fragments created by enzymes such as Notl, since some of these fragments will be very long. Also, the choice between different possible end points that is available in a partial digest library is lost in complete digest libraries, making them more sensitive to repetitive or unclonable se- quences. Some of these problems can be overcome in a jumping library constructed from DNA fragments gener- ated by digesting long DNA partially with an enzyme cutting commonly (e.g. EcoR]) and completely with an enzyme cutting very rarely in the genome (e.g. Notl) and circuJArizing over a plasmid cleaved with beth enzymes. Such a library will both preserve the advan- tage of the small number of Notl sites constituting one of the end points and allow rmmy more possible start points, if screened by fragments adjacent toEcoRI sites.

Using these strategies and a number of newly con- structed vectors, we have constructed several jumping libraries. Those recently completed and currently being tested are a human NotllBamHl library, a human Notl complete-EcoRI partiallBamHI re-cleaved library, and mouse NotIIBamHI and a MlullEcoRI libraries. Ran- domly picked clones from these libraries in all cases give the expected restriction digest patterns. In addition, for a number of clones picked rondomly from the human NotllBamHl library (containing 100000 clones corres- ponding to 10--30 genomes) both end points of the clone have been shown to hybridize to Notl fragments of i~entical size (Poustlm et al., unpublished). Work on the identification and analysis of specific clones is in progress.

Linking libraries If a jumping library is constructed with an enzyme that

cuts rarely in the genome, a strategy is needed for making ordered jumps through the library. It should in theory be possible to do this by making a complementary library, as described above and in Figs 2 and 3. However it is likely that the effidency of constructing such a library will be low because only a small fraction of circularized parfial BamHl digestion products recleaved withNotl will be within a clonable size range.

• perspect ives

• . / / Chromosome

=. v / , __4 : Linking library

Fig. 5. D i ~ jumping by alternating between rare cutter j u m p i ~ libra~ and complementa~ linking library.

There are a number of possible sol~'ons to the free host, and clones from the human chromosome problem. A fragment for the next step can be provided fragments are identified by hybridization to human by a ~, or cosmid clone isolated by using the jumping- repetitive DNA probes ix. By use of this protocol a library done as a probe. Alternatively, special 'linking number of clones have been isolated, characterized by libraries', containing only genomic sequences surround- restriction analysis and mapped to either chromosome 4 ing, for example, the Notl sites, can be constructed and or chromosome 5 by hybridization to appropriate so- screened (Figs 4 and 5). Each clone isolated from such a matic cell hybrids. Two clones located on chromosome 4 Nofl linking library will overlap two ne~ghbouring Notl have been analysed further by hybridization to N~t l restriction fragments, which can be identL~l by hyb~d- digests of human DNA separated by pulsed field gradient izaflon of the done or sub-fragments to appropriate gel electrophoresis. As expected, each clone can be pulsed field gradient gels. Therefore, if chromosome shown to hybridize to two separate Notl fragment regions short enough to comprise only a smallnumber of bands. distinguishable Notl restriction fragments are saturated Other techniques likely to be useful in the construc- with linking clones, the order of the clones can be tion of linking libraries are also being developed. For established (Fig. 6). This method allows combined instance, we have constructed aNotlinsertion vector to physical and genetic maps to be made of regions enable us to use, for example, complete EcoRl digests surrounding specific mutations, and eventually maps of derived from either sorted chromosomes or chromo- chromosomes and even entire genomes can be obtained, some h"oraries as starting material. Also, a number of

One strategy for cloning Nofl linking fragments plasmids carrying polylinkers with rare cutting sites (defined by the subset of Notl sites cut in tit.., genomic 0Votl, Mlul, Sacll) have been constructed (but not yet DNA) is shown on the fight-hand side of Fig. 4. We are testedin this application) with the aim of allowing cosmid using this strategy, in collaboration with the labomtories or ~. c|~nes carrying these sites to be selected from of Gusella, Cantor and Wasmuth, to saturate a regionin pre-existing h"oraries. the vicinity of the Huntington's chorea locus (Frischauf and Pohl, unpublished). Teehnicalproblems

In a first step, DNA from a hamster cell line, In both the construction and screening of complete containing parts of the human chromosomes 5 and 4, jumping libraries there are a number of complications including the Huntington's chorea locus, is partially that are not encountered in the construction of other cleaved with Sau3A and si¢ed on sucrose gradients to types of libraries. Difficulties include the problems of isolate fragments of 10-20 kb. This DNA is then generating andhandlingthe very large DNAmolecules circularized at low concentration in the presence of a required, without physical shearing or enzymatic de- BamHl-cut plasmid containing a suppressor gene s gradation taking place, the powerful genetic or physical re-cleaved with Notl, and ligated into Notl-cleaved selection required to select the subset ofdones contain- NotEMBL3AtZ, an analogue of the ~ replacement vector ing the tag sequence, and the large amounts of material EMBL3A carryingNotlsitesin the polylinkersequence, that in most cases have to be processed to generate Linking clones are selected by plating on a suppressor- complete libraries.

I m

' , " ¢1 - - - ' I PFG pattern

Linking clones

hamster human hamster Fig. 6. Detenniaing the order of linking clones in a region of the hman genome by IObtidi='ng the clones to pulsed. ~ld gradient gels and identifying where two clones hyb~d~ to a common res~.fragnm~.~ In this way a restriction map can be establisiw.d over a large gene~ distance.

177

perspectives For e.,,~mple, in a representative BamHl/EcoRl lib-

rary construction ~jumping dis~nce 200 kb, expected ratio ofintra- to intermolecular ligation 10: 1) circulafiza~ tion ligations of 20--40 ml (10--20 ~g DNA at a concen- tration of 0.5 lag nd -l or less) have to be carded out. After cleavage with the second enzyme, the DNA fragments have to be ligated with 200-400 ttg of ~, vector to ensure molar excess of the vector, because only a small fraction (typically less than 1%) of the very large number of DNA fragments created during the second digestion step will carry the tag sequence. Packaging reactions and plating have to be scaled up accordingly. Even neglecting the particular problems in handling large DNA molecules, construction of jumping libraries is therefore much more difficult than, for instance, the construction of a cosmid library of equivalently complete coverage.

As mentioned before, the work can be reduced considerably if enzymes cutting only rarely in the genome are used. This, however, does require an increased effort in finding useful starting points (e.g. Notl, Mlul, BssHII sites) within or close to the available marker fragments. This task is further complicated by the sensitivity of these enzymes to modification of the genomic DNA (presumably methylation of the CG sequences contained within the recognition sites), lead- ing to incomplete digestion (Barlow and Bucan, unpub- lished).

Persistent problems with suppressor-independent phages arising by recombination with the packaging phage have been encountered, but were solved in our laboratory by the construction and use of special phage vectors with multiple amber mutations (so-called 'Aam- BamSam' vectors) for the construction of jumping libraries.

While we are fairly confident that jumping libraries aiming at jumps of a few hundred kilohases w~l predomi- nantly contain the products of intrarnolecular ligations, it can be expected that reduced ligation e~ciencies and higher probabilities of random breakage will limit the size of the jumps in ck,nes which can be recovered at reasonable frequency.

Alternative strategies Although a number of essentially complete jumping

and linking libraries have already been constructed and screened, alternative strategies are being investigated to overcome some of the remaining difficulties and limitations of the technique.

To reduce the amount of DNA handled in the vector ligation, packaging and plating, it might be advantageous if genetic selection for the clones containing the tag sequence could be replaced and/or complemented by physical isolation protocols. Procedures considered 3 include the use of tag sequences carrying biotin- substituted nucleotides, which could be selected by binding to avidin columns, or the use of/ac operator sequences as tags, which can be isolated by their specific interaction with lac repressor protein.

To overcome the need for screening the very complex libraries which, depending on the enzymes used, will be required to cover the entire genome, either genetic selection protocols 13,14, or physical selection pro- cedures could be used. To allow the application of such an alternative physical selection protocol, the plasmids used in library construction could be packaged in M13 phage particles as single stranded circular DNA 178

TIG - - Ju~ 1986

molecules 7's. Specific clones could then be enriched by hybridization, and recovered by transformation into E. coll. Since genetic selection based on homologous r~mbination in E. coil ~ t e s s~c,~-~y a~'~t sequence differences found between different copies of repeated sequences Is, the implementation of genetic selection protocols might also reduce the difficulties caused by repeated sequences.

To increase the efficiency of library construction, it might be useful to develop strategies in which either no tag sequences are required (possible for complete digest libraries), or in which !igation of the tag (to one end only) is carried out before the circularization reaction. As a test of this possibility, a tag plasmid cleaved both with Notl and EcoRI and treated with phosphatase, was ligated at high DNA concentrations to DNA digested completely with Notl. The ligation product was then re-cleaved partially withEcoRi, diluted and circularized. So far, clones have been recovered at a frequency sinfilar to that found when the standard protocol was used, but we anticipate further improvements.

As a possible route to increase the w ~ m i size of jumps, ligation-independent circularization strategies might be explored. These could involve either circular- ization by hybridization of complementary single stranded ends provided by the tags, or the use of in vitro recombination between tag sequences camJin" g recog- nition sites for site-specific recombinases le.

Prospects We are at a time when the technical capabilities for

handling, cloning and analysing the large physical dis- tances characteristic of mammalian genomes are being developed rapidly. These developments coincide with rapid progress in mamwaHa_n genetics, made possible by the use of DNA probes as highly polymorphic markers l~ to identify the genetic and physical environment of mammalian mutations. The combination of high resolu- tion genetics with long range cloning and map,ulng techniques can be expected to allow, at least in some cases, the identification and molecular analysis of genes identified by such mutations.

We are testing these approaches in an analysis of regions of mouse chromosome 17 to localize physically a number of genes implicated in embryonic development (e.g. T, Fu, t haplotype lethals) or other genetically defined effects associated with this region. Simil~ly, we and others are attempting to use the techniques in the analysis of the environment of human mutations that have been identified as hereditary diseases and for which linked markers have become available (e.g. Hunting- ton's chorea, cystic fibrosis, Duchenne muscular dys- trophy).

Progress in this rapidly developing field of cloning and DNA analysis techniques may therefore prove to have major implications for many areas of mammalian genetics and medicine.

Acknowledgements We thank Francis Collins, Sherman Weissman,

Charles Cantor and Cassandra Smith for unpublished information on work on jumping and linking libraries respectively, and Anna-Maria Frischauf, Thomas Pohl, Denise Barlow, Lisa Stubbs, Margit Burmeister and Elliot Ehrich for experimental results mentioned in the review.

TIG - - J u l y 198~q

References I Smith, C. L., Lawrance, S. K., Gillespie, G. A., Cantor, C. R.,

W ~ x m , S. M. and ColOns, F. S. Methods Emzymol. (in press) 2 Wyman, A_ R., Wolfe, L. B. and Botstein, D. (1985) Proc. Nail

Am& SoL USA 82, 288O-2884 3 Coins, F. S. and Weissman, S. M. (1984) Proe. N~Acad. Sd.

U&4 81, 6812-6816 4 Marx, J. L. ( 1 9 8 5 ) ~ ZZS, 1080 5 S~u"dz, C. L. and Cantor. C. IL Methods Enzymol. (in press) 6 Schwartz, D. C. and Creator, C, R. (1984) Cell 37, 67-75 ? Levinson, A., Silver, D. and Seed, B. (1984)]. Mol. AI~I. Greet

2~ 507-517 8 Hmng, H. V., Little, P. F. R. andSeed, B. inVedors, aSur~o f

Molecular Cio. i~ Vectors and their Uses (Rodriguez, R., e(L), Butterworth Press (in press)

9 Jacobson, EL and Stockmayer, W. H. (1950) ]. C/-.~'m. P~s. 18, 1600-1606

reviews 10 Clarke, L, and Carbon, J. (1976) Cell 9, 91-99 11 Crampton, J. M., Davies, K. E. and Fmapp, T. F. (1981) Nuc/e/¢

Adds Res. 9, 3821--3833 12 Frischanf, AM., Murray, N. and Lehrach, H. Methods E~m~ol.

(in press) 13 Seed, B. (1983)N~de/c Adds Res. 11, 2427-2445 14 Poustka, A., Packwitz, HR., Frischauf, AM., Hohn, B. and

Lehraclt H. (1984)Proc. NaIIAcad. Sd. USA 81, 4129.-4133 15 Shen, P. and Huang, H. V. (1986) Gmet/cs 112, 441-457 16 Abremsld, A. and Hoess, R.H. (1984)]. Biol. Chem. 259,

1509-1514 !7 Botste~ D., White, R. L., Skolnick, M. and Davis, R. W. (1980)

Am. ]. Hum. Genet. 32, 314--331

A. Poustka and H. Lehrach are at the European Molecular - B~?o~ Laborakny, D.69, Heidelberg, Meyerhofstr. 1, FRG.

T lymphocytes are involved in a vm~ety of immunological (regu- latory and effector) functions, such as provision of help for antibody formation by B cells and ldlling of virally infected cells. In addition to the T cell receptor for antigen (TCR), a number of additional T cell- specific surface proteins play important roles in these func- tions. These molecules have been referred to as T cell differentiation antigens because they are expressed at defined times during T cell development and, in some cases, on specific functional subsets. Four of these molecules will be discussed here" two which are involved in T cell activation pathways (CD3 and CD2) and two which define major subsets of peripheral T cells and are linked to the T cell recognition process (CD8 and CD4). Although this review deals primarily with the human proteins, analogies and contrasts with the mouse equivalents are also discussed.

Structure and function of T lymphocyte differentiation

antigens Jane R. Parnes

In addition to the T cell recOttn for anligen, T lymphocytes express a variely of addi6onal T cell-specific surface molecub~s which are involved in reco~zition, activation and func6on. The stmdure, evolWionary relationshO and functions of four of these T cell

differentiation antigens are discussed.

The CD3 molecule is associated with the T cell antigen receptor

Three T cell-specific protein chains which are non-covalently associated with one another form the human CD3 molecule (also called 1"3 or Leu-4): ? (25 kDa), 6 (20 kDa) and c (20 kDa). All three chains appear to be invariant. The • chain is highly hydrophobic and non-glycosylated, in contrast to the other two chains. A cDNA clone for the 6 chain has been isolated and predicts a mature protein of 150 amino acids with 79 extracellular residues, 27 within the membrane and 44 within the cytoplasm ~. It is not homologous tc other known proteins and is encoded by a single gene.

The CD3 molecule is expressed on some thymocytes and on all mature T cells. It is closely associated with the TCR, as. demonstrated by the finding that monoclonal antibodies (Mab) specific for either CD3 or the TCR comodulate both from the cell surface or immunoprecipi- tate both z'3. Chemical cross-linking data show that the interaction is primarily between the [3 chain of the TCR and the CD3 y chain, and suggest a stoichiometry of I: 1 (Ref. 4). CD3 is only expressed on _the T cell surface in the presence of the two chains of the TCR (oh and [i) (Ref. 5), but it is not yet known whether surface expression of the TCR requires the presence of CD3. ~) 1986, Elsevier Science Pub~hers B.V.. Amsterdam 0168 - 9525/86/$0.200

Although there are no Mab against a mouse eqt~valent of CD3, a similar set of TCR-associated polypeptides has been identilied s'e. However, the pattern in the mouse appears more complex, with at least four or five distinct polypeptide chains.

The precise function of CD3 is not known, but clues have been derived from the functional effects of Mab specific for this protein. Such Mab inhibit antigen-driven T cell proliferation, presumably because of m~uiation of the TCR--CD3 complex from the cell surface z'~. They also block killing by cytotoxic T cells (CTL) at a step beyond the recognition phase, and possibly involving the triggering of the lethal hit z'7's. T cell clones show enhanced responsiveness to the proliferative effects of the T cell growth factor interleukin-2 (IL-2) and an increase in the number of IL-2 receptors after treatment with soluble anti-CD3 Mab 9. In the presence of a secona signal, which can be provided by accessory cells (monocytes or macrophages) or exogenous interleu- kin-1 (I_Lol), Sepharose-bound anti-CD3 Mab induce activation of resting T cells, as measured by proliferation or lymphokine release (IL-2 or v-interferon) 7'9. In the case of T cell clones, no second signal is needed to induce activation 9. Sepharose-bound Mab specific for the TCR @milady activate T cells, and in both cases the bound antibodies are thought to mimic activation induced by antigen (Fig. 1).

Exposure of T cell clones to Sevharose-bound anti-CD3 or anti-TCR Mab, or ~i cell tumors to soluble anti-CD3 or anti-TCR Mab is followed within minutes by a rapid rise in the intracellular concentration of free Ca 2+ (Refs 10-12). In T cell clones and the HPB-ALL tumor line this increase appears to result from an influx of Ca 2+ through a plasma membrane channel ma2, but in the tumor line Jurkat, a release from intraceilular stores has

179